Stock Search POC

Filters

Search Mode

Vector only (semantic)

Hybrid (text + vector RRF)

Linear (text + vector weighted)

RRF Rank Constant

Controls influence of lower-ranked documents in RRF fusion. Lower values give more weight to top results. Default: 60

Retriever Weights:

Text Retriever Weight: 1.0

0.02.55.0

Weight for text search over title and tags fields. Default: 1.0

Vector Retriever Weight: 1.0

0.02.55.0

Weight for vector search over embedding_text field. Default: 1.0

Higher weights increase the influence of that retriever in the final RRF ranking. Both retrievers contribute to results, but their relative weights determine how much each affects the final score.

Hybrid mode combines text search (title + tags) with vector search using Reciprocal Rank Fusion (RRF).

Retriever Weights:

Text Retriever Weight: 1.0

0.02.55.0

Weight for text search over title and tags fields. Default: 1.0

Vector Retriever Weight: 1.0

0.02.55.0

Weight for vector search over embedding field. Default: 1.0

Normalizer

Normalization method for score combination.

Higher weights increase the influence of that retriever in the final linear ranking.

Linear mode combines text search (title + tags) with vector search using weighted linear combination with normalization.

Embedding Model

Google Vertex AI (multimodal)

TwelveLabs Marengo 3.0

Uses TwelveLabs Marengo 3.0 embeddings. Searches the embedding_twlv field.

AI Prompt Enhancement

Enhance query with AI

Number of Prompts (N):10

Number of distinct queries to generate (1-20)

AI Results Mix:60% AI / 40% Raw

0%50%100%

Percentage of results from AI-generated queries vs raw query

Word Threshold:5 words

Enhance queries shorter than this many words (1-20)

System Prompt:

You are an intelligent Search Expansion Engine for a professional stock asset p...

Customize the AI prompt template (must include {user_query} and {N} placeholders)

Score Dampening

Enable Score Dampening

Dampening Factor:0.95

More Diversity (0.0)More Relevance (1.0)

Lower values increase diversity by penalizing consecutive results from the same query. Higher values maintain more relevance to the original scores.

Jina Reranking

Enable Jina Reranking

Jina Weight Factor

Weight factor for adding Jina scores to Elasticsearch scores. Formula: Combined Score = ES Score + (Weight × Jina Score). Default: 0.5

Result Diversification

Enable Result Diversification

Lambda: 1.00Query similarity

0.00.51.0

*Currently not supported by the es version. Controls similarity vs diversity. Higher values (closer to 1.0) weight similarity to the query more heavily. Lower values (closer to 0.0) weight diversity more heavily. Default: 1.0 (no diversification).

Result Biasing

Enable Result Biasing

Intelligent result mixing using targeted Elasticsearch queries to guarantee minimums and respect maximums for different asset types.

Max Transparent Results (%)

Maximum percentage of transparent assets allowed in results. Used as an upper limit during result mixing (0-100)

Min Non-Transparent (%)

Guaranteed minimum percentage of non-transparent, non-video assets. Uses targeted search queries to ensure reliability. Videos are excluded from this calculation (0-100)

Transparent Penalty (0.0-1.0)

Score multiplier applied to transparent assets after fetching. Lower values reduce their ranking priority (0.0-1.0)

Video Penalty (0.0-1.0)

Score multiplier applied to video assets after fetching. Lower values reduce their ranking priority (0.0-1.0)

Asset Type

All

Images

Videos

Templates

Subcategory

All

PhotosIllustrationsVectors

Price

All

Free

Standard Content

Premium and video

Usage Rights

All

Commercial Use

Editorial Use

Background

All

TransparentIsolated Assets

Generative AI

All

Generative AI Only

Exclude Generative AI

Icons

Icon SheetIndividual Icon

Orientation

All

Horiz.VerticalSquarePanoramic

Size

px

px

MP

Creation Date

Content Country

Keywords Language

Default Settings:

Search Mode:Vector Only

Embedding Model:TwelveLabs

AI Enhancement:Enabled

Jina Reranking:Enabled

Score Dampening:Enabled

Start your search

Enter a query above to search through Adobe Stock assets

Press Enter to search